CN109800302A - Public sentiment method for early warning, device, terminal and medium based on Recognition with Recurrent Neural Network algorithm - Google Patents
Public sentiment method for early warning, device, terminal and medium based on Recognition with Recurrent Neural Network algorithm Download PDFInfo
- Publication number
- CN109800302A CN109800302A CN201811530781.XA CN201811530781A CN109800302A CN 109800302 A CN109800302 A CN 109800302A CN 201811530781 A CN201811530781 A CN 201811530781A CN 109800302 A CN109800302 A CN 109800302A
- Authority
- CN
- China
- Prior art keywords
- keyword
- public sentiment
- correlation
- news
- recognition
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention belongs to nerual network technique fields, disclose a kind of public sentiment method for early warning based on Recognition with Recurrent Neural Network algorithm, device, terminal and medium, by obtaining public sentiment news in preset time, and determine the tendency degree of keyword in the public sentiment news, further according to the tendency degree of the keyword, determine the corresponding feature vector of the keyword, further according to the corresponding feature vector of the keyword, determine the characteristic sequence of the public sentiment news, the characteristic sequence of the public sentiment news is finally inputted to the Recognition with Recurrent Neural Network model trained, determine public sentiment warning index, according to the public sentiment warning index, issue public sentiment early warning, it can be moved towards with accurate judgement public sentiment, solves the technical problem of the prediction of the development trend effect difference of prior art public sentiment.
Description
Technical field
The present invention relates to field of artificial intelligence more particularly to a kind of public sentiment early warning based on Recognition with Recurrent Neural Network algorithm
Method, apparatus, terminal and medium.
Background technique
With the fast development of Internet technology, the exploration and flexibility of network allow it to become the master of reflection social public opinion
Want one of carrier.Public sentiment early warning can find relevant to " I " public feelings information, negative information in first time, great public sentiment and
When early warning；The analysis of public opinion data of qualitative, quantitative, the development and change of the specific public sentiment of accurate judgement or a certain public sentiment special topic are provided
Trend；Public sentiment report and various statistical report forms are automatically generated, the quality and efficiency of public sentiment management are improved, assist managerial decision.
Currently, there are many public sentiment method for early warning in the market, but there are many deficiencies and defect, such as the development of public sentiment
Trend prediction effect is poor.
Above content is only used to facilitate the understanding of the technical scheme, and is not represented and is recognized that above content is existing skill
Art.
Summary of the invention
The main purpose of the present invention is to provide a kind of public sentiment method for early warning based on Recognition with Recurrent Neural Network algorithm, dress
It sets, terminal and medium, it is intended to solve the technical problem of the prediction of the development trend effect difference of prior art public sentiment.
To achieve the above object, the present invention provides a kind of public sentiment method for early warning based on Recognition with Recurrent Neural Network algorithm,
It is characterized in that, includes the following steps:
Public sentiment news in preset time is obtained, and determines the tendency degree of keyword in the public sentiment news；
According to the tendency degree of the keyword, the corresponding feature vector of the keyword is determined；
According to the corresponding feature vector of the keyword, the characteristic sequence of the public sentiment news is determined；
The characteristic sequence of the public sentiment news is inputted to the Recognition with Recurrent Neural Network model trained, determines that public sentiment early warning refers to
Mark；
According to the public sentiment warning index, public sentiment early warning is issued.
Preferably, public sentiment news in the acquisition preset time, and determine the tendency degree of keyword in the public sentiment news
The step of, comprising:
Public sentiment news and the keywords database pre-established in preset time are obtained, and is determined crucial in the public sentiment news
The tendency degree of word.
Preferably, the tendency degree of the keyword includes positive tendency degree, negative tendency degree and neutral tendency degree, described
Positive tendency degree, negative tendency degree and neutral tendency degree for the keyword respectively appear in positive news, negative press, in
Probability in vertical news；
Correspondingly, public sentiment news and the keywords database pre-established in the acquisition preset time, and determine the carriage
Before the step of tendency of keyword is spent in feelings news, the public sentiment method for early warning based on Recognition with Recurrent Neural Network algorithm further includes
Following steps:
Keywords database is established, the keywords database includes positive keyword set, negative keyword set and neutral keyword
Collection；
Calculate the correlation of each keyword with remaining keyword in each keyword set；
According to the correlation of each keyword and remaining keyword in each keyword set, the front for calculating the keyword is inclined
Xiang Du, negative tendency degree and neutral tendency degree.
Preferably, the correlation according to each keyword and remaining keyword in each keyword set, calculates the pass
The step of positive tendency degree, negative tendency degree and the neutral tendency of keyword are spent, comprising:
Keyword is closed with the correlation of remaining keyword in positive keyword set, with remaining in negative keyword set
The correlation of keyword, the positive tendency degree of difference conduct with the correlation of remaining keyword in neutral keyword set；
Keyword is closed with the correlation of remaining keyword in negative keyword set, with remaining in positive keyword set
The correlation of keyword, with the difference of the correlation of remaining keyword in neutral keyword set as negative tendency degree；
Keyword is closed with the correlation of remaining keyword in neutral keyword set, with remaining in positive keyword set
The correlation of keyword, the neutral tendency degree of difference conduct with the correlation of remaining keyword in negative keyword set.
Preferably, the step of correlation for calculating each keyword and remaining keyword in each keyword set, packet
It includes:
According to formulaRemaining calculated in the keyword and positive keyword set is crucial
The correlation of word；
Wherein, n is the number of keyword in positive keyword set；
Rec (w, v) is w, the correlation of two keywords of v；
P is positive keyword set；
P (w) is the probability that w keyword occurs in a document,
P (v) is the probability that v keyword occurs in a document；
P (w, v) is the probability that w and v occur in a document jointly.
Preferably, the step of correlation for calculating each keyword and remaining keyword in each keyword set, packet
It includes:
According to formulaCalculate remaining pass in the keyword and negative keyword set
The correlation of keyword；
Wherein, m is the number of keyword in negative keyword set；
Rec (w, v) is w, the correlation of two keywords of v；
Q is positive keyword set；
P (w) is the probability that w keyword occurs in a document,
P (v) is the probability that v keyword occurs in a document；
P (w, v) is the probability that w and v occur in a document jointly.
Preferably, the step of correlation for calculating each keyword and remaining keyword in each keyword set, packet
It includes:
According to formulaRemaining calculated in the keyword and neutral keyword set is crucial
The correlation of word；
Wherein, k is the number of keyword in neutral keyword set；
Rec (w, v) is w, the correlation of two keywords of v；
M is neutral keyword set；
P (w) is the probability that w keyword occurs in a document,
P (v) is the probability that v keyword occurs in a document；
P (w, v) is the probability that w and v occur in a document jointly.
Based on foregoing invention purpose, the present invention also provides a kind of public sentiment prior-warning device based on Recognition with Recurrent Neural Network algorithm,
Include:
Public sentiment obtains module, for obtaining public sentiment news in preset time, and determines keyword in the public sentiment news
Tendency degree；
Vector establishes module, for the tendency degree according to the keyword, determines the corresponding feature vector of the keyword；
Sequence determining module, for determining the feature of the public sentiment news according to the corresponding feature vector of the keyword
Sequence；
Index determining module, for the characteristic sequence of the public sentiment news to be inputted the Recognition with Recurrent Neural Network mould trained
Type determines public sentiment warning index；
Early warning issues module, for issuing public sentiment early warning according to the public sentiment warning index.
Based on foregoing invention purpose, the present invention also provides a kind of terminal, the terminal includes: memory, processor and deposits
Store up the public sentiment early warning program based on Recognition with Recurrent Neural Network algorithm that can be run on the memory and on the processor, institute
The public sentiment early warning program based on Recognition with Recurrent Neural Network algorithm is stated to be arranged for carrying out such as the above-mentioned algorithm based on Recognition with Recurrent Neural Network
The step of public sentiment method for early warning.
Based on foregoing invention purpose, the present invention also provides a kind of storage medium, it is stored on the storage medium and is based on following
The public sentiment early warning program of ring neural network algorithm, the public sentiment early warning program based on Recognition with Recurrent Neural Network algorithm are held by processor
It realizes when row such as the step of the above-mentioned public sentiment method for early warning based on Recognition with Recurrent Neural Network algorithm.
The present invention determines the tendency degree of keyword in the public sentiment news by obtaining public sentiment news in preset time,
Further according to the tendency degree of the keyword, the corresponding feature vector of the keyword is determined, it is corresponding further according to the keyword
Feature vector determines the characteristic sequence of the public sentiment news, has finally trained the characteristic sequence input of the public sentiment news
Recognition with Recurrent Neural Network model determines public sentiment warning index, according to the public sentiment warning index, issues public sentiment early warning, can be accurate
Judge that public sentiment is moved towards, solves the technical problem of the prediction of the development trend effect difference of prior art public sentiment.
Detailed description of the invention
Fig. 1 is the structural schematic diagram of the terminal for the hardware running environment that the embodiment of the present invention is related to；
Fig. 2 is that the present invention is based on the flow diagrams of the public sentiment method for early warning first embodiment of Recognition with Recurrent Neural Network algorithm；
Fig. 3 is that the present invention is based on the flow diagrams of the public sentiment method for early warning second embodiment of Recognition with Recurrent Neural Network algorithm；
Fig. 4 is that the present invention is based on the flow diagrams of the public sentiment method for early warning 3rd embodiment of Recognition with Recurrent Neural Network algorithm；
Fig. 5 is that the present invention is based on the flow diagrams of the public sentiment method for early warning fourth embodiment of Recognition with Recurrent Neural Network algorithm；
Fig. 6 is that the present invention is based on the structural block diagrams of the public sentiment prior-warning device first embodiment of Recognition with Recurrent Neural Network algorithm.
The embodiments will be further described with reference to the accompanying drawings for the realization, the function and the advantages of the object of the present invention.
Specific embodiment
It should be appreciated that described herein, specific examples are only used to explain the present invention, is not intended to limit the present invention.
Referring to Fig.1, Fig. 1 is the terminal structure schematic diagram for the hardware running environment that the embodiment of the present invention is related to.
As shown in Figure 1, the terminal may include: processor 1001, such as central processing unit (Central Processing
Unit, CPU), communication bus 1002, user interface 1003, network interface 1004, memory 1005.Wherein, communication bus 1002
For realizing the connection communication between these components.User interface 1003 may include display screen (Display), input module ratio
Such as keyboard (Keyboard), optional user interface 1003 can also include standard wireline interface and wireless interface.Network interface
1004 may include optionally standard wireline interface and wireless interface (such as Wireless Fidelity (WIreless-FIdelity, WI-FI)
Interface).Memory 1005 can be random access memory (Random Access Memory, RAM) memory of high speed,
It can be stable nonvolatile memory (Non-Volatile Memory, NVM), such as magnetic disk storage.Memory 1005
It optionally can also be the storage device independently of aforementioned processor 1001.
It will be understood by those skilled in the art that the restriction of the not structure paired terminal of structure shown in Fig. 1, may include ratio
More or fewer components are illustrated, certain components or different component layouts are perhaps combined.
As shown in Figure 1, as may include operating system, data storage mould in a kind of memory 1005 of storage medium
Block, network communication module, Subscriber Interface Module SIM and the public sentiment early warning program based on Recognition with Recurrent Neural Network algorithm.
In terminal shown in Fig. 1, network interface 1004 is mainly used for carrying out data communication with network server；User connects
Mouth 1003 is mainly used for and user carries out data interaction；Processor 1001, memory 1005 in terminal of the present invention can be set
In the terminal, the terminal calls the carriage based on Recognition with Recurrent Neural Network algorithm stored in memory 1005 by processor 1001
Feelings early warning program, and execute the public sentiment method for early warning provided in an embodiment of the present invention based on Recognition with Recurrent Neural Network algorithm.
The embodiment of the invention provides a kind of public sentiment method for early warning based on Recognition with Recurrent Neural Network algorithm, referring to Fig. 2, Fig. 2
For the present invention is based on the flow diagrams of the public sentiment method for early warning first embodiment of Recognition with Recurrent Neural Network algorithm.
In the present embodiment, the public sentiment method for early warning based on Recognition with Recurrent Neural Network algorithm includes the following steps:
Step S10: public sentiment news in preset time is obtained, and determines the tendency degree of keyword in the public sentiment news；
It should be noted that the executing subject of the present embodiment method is terminal, public sentiment news is a kind of network public-opinion, is passed through
The network platform is diffused and propagates.Public sentiment news can be through the publications such as webpage or third party software, plug-in unit.And
The acquisition of public sentiment news can be to be obtained by api interface, is also possible to obtain by modes such as spiders, is not done have herein
Body limitation.
The tendency degree of keyword can be divided into positive tendency degree, negative tendency degree, can also be divided into positive tendency degree, negatively
Tendency degree and neutral tendency degree.The positive tendency degree of keyword is the degree of front evaluation, and the negative tendency degree of keyword is
The degree of unfavorable ratings, the neutral tendency degree of keyword are the degree of neutral evaluation.
Before the public sentiment news in acquisition preset time, need to pre-process public sentiment news, pretreated method
Include:
Step S100a: public sentiment news is clustered.Due to public sentiment news have in time it is sudden, do not have it is general
Time rule, it is therefore desirable to (such as leave office about company executives occurrences in human life, corporate strategy policy phase to the same topic of public sentiment news
Inside the Pass hold etc.) quantity predicted.Cluster process is mainly by the described public sentiment news Aggreagation for same topic to same
In classification.Clustering method can be not particularly limited herein using clustering method conventional in the prior art.
Step S100b: associated topic is obtained.The public sentiment news quantity occurred on network is many by meeting, corresponding topic
It also can be very much.The topic that is generally concerned with of public sentiment prediction can be that user is customized, be also possible to be set as the routine of enterprises pay attention
Topic, such as the leaving office of company executives occurrences in human life, corporate strategy policy related content etc..Associated topic in public sentiment news is obtained, it can be with
It is to be obtained by keyword retrieval, other conventional means can also be used, be not particularly limited herein.
Step S100c: data aggregate is carried out to public sentiment news.By carrying out data aggregate to public sentiment news, one is obtained
Time series, the value at each moment are the quantity of all public sentiment news on network until current time.
When specific implementation, public sentiment news in preset time is obtained, is divided using public sentiment news of the participle tool to acquisition
Word obtains the keyword in every public sentiment news, then determines the tendency degree of keyword.
It determines that the tendency degree of keyword can be acquisition history public sentiment news in advance, history public sentiment news is marked,
Go out in the public sentiment news of keyword occurs in the public sentiment news of front evaluation in statistics public sentiment news number or unfavorable ratings
Existing number establishes the tendency degree library of keyword with this；When the tendency for determining some keyword is spent, then looked into tendency degree library
Look for the corresponding tendency degree of keyword.
Step S20: according to the tendency degree of the keyword, the corresponding feature vector of the keyword is determined；
It should be noted that the tendency degree according to the keyword, determines the corresponding feature vector of the keyword
It is to construct the corresponding feature vector of keyword using the tendency degree of keyword as corresponding weight.The dimension of feature vector can be with
Depending on the division of tendency degree, such as tendency degree is divided into positive tendency degree, negative tendency degree and neutral tendency degree, then feature
Vector can be set at least three-dimensional.
Step S30: according to the corresponding feature vector of the keyword, the characteristic sequence of the public sentiment news is determined；
It should be understood that public sentiment news be by multiple crucial phrases at, it is described according to the corresponding spy of the keyword
The step of levying vector, determining the characteristic sequence of the public sentiment news can be that the corresponding combination of eigenvectors of keyword is formed carriage
The characteristic sequence of feelings news.For example, keyword is m in a public sentiment news, the characteristic sequence for constructing the public sentiment news can be with
It, can also be according to the dimension of specific determining characteristic sequence of classifying for the dimensional feature vector of 3 × m or m × 3.
Step S40: the characteristic sequence of the public sentiment news is inputted to the Recognition with Recurrent Neural Network model trained, determines public sentiment
Warning index；
It should be understood that using the characteristic sequence of public sentiment news as the input of Recognition with Recurrent Neural Network model, circulation nerve
The hidden layer of network includes the implicit vector of history public sentiment news, and by Recognition with Recurrent Neural Network model, the entirety for obtaining public sentiment is inclined
Xiang Du, as public sentiment warning index.
In addition, since public sentiment news is at a time or in the period, the regional range that public sentiment news is influenced is also
It is different, such as country, province, city etc., thus consider public sentiment news quantity be can be by the regional range of public sentiment news
As a correction value, it is denoted as t1.
Since the report of media also can have large effect to the quantity of public sentiment news, the exposure of public sentiment news
Degree is also considered as a correction value, is denoted as t2.
The circulation of public sentiment news also can reflect out public feelings information propagation condition and discussion temperature on network, because
This, the circulation of public sentiment news can also be used as a correction value, be denoted as t3.It is customized that the setting of correction value t3 can be user.
Such as the circulation possibility of the public sentiment news about the leaving office of company executives occurrences in human life might not be big, but for enterprise, this
The attention rate of one topic is often relatively high, therefore can be modified by adjusting t3.
The training method of Recognition with Recurrent Neural Network model, which can be, grabs public sentiment news data by network；Loop initialization mind
Parameter through network model calculates circulation nerve according to keyword data in public sentiment news data and Recognition with Recurrent Neural Network model
Parameter in network model, specific training method can also adopt with method known in this field.
Step S50: according to the public sentiment warning index, public sentiment early warning is issued.
It is described according to the public sentiment warning index when specific implementation, public sentiment early warning is issued, can be in public sentiment warning index
When greater than a preset threshold, public sentiment early warning is issued.Wherein it is customized to can be user for preset threshold, can also be new according to public sentiment
The topic content of news is preset.The mode for issuing public sentiment early warning can be a variety of, such as pass through mobile portable phone, short message, mail
Etc. forms, be also possible to the customized simultaneous system of user.
The present invention determines the tendency degree of keyword in the public sentiment news by obtaining public sentiment news in preset time,
Further according to the tendency degree of the keyword, the corresponding feature vector of the keyword is determined, it is corresponding further according to the keyword
Feature vector determines the characteristic sequence of the public sentiment news, has finally trained the characteristic sequence input of the public sentiment news
Recognition with Recurrent Neural Network model determines public sentiment warning index, according to the public sentiment warning index, issues public sentiment early warning, can be accurate
Judge that public sentiment is moved towards, solves the technical problem of the prediction of the development trend effect difference of prior art public sentiment.
It is that the present invention is based on the processes of the public sentiment method for early warning second embodiment of Recognition with Recurrent Neural Network algorithm with reference to Fig. 3, Fig. 3
Schematic diagram.
Based on above-mentioned first embodiment, in the present embodiment, the step S10, comprising:
Step S101 obtains public sentiment news and the keywords database pre-established in preset time, and determines the public sentiment
The tendency degree of keyword in news.
It should be noted that keywords database can also be divided into positive keyword set, negative keyword set and neutral key
Word set, or it is divided into front keyword and negative keyword set, specific mode classification is set according to demand.
The keywords database pre-established can be the public sentiment news according to tape label, be that the public sentiment that front is evaluated is new by label
The keyword occurred in news is put into positive keyword set, and label is put for the keyword occurred in the public sentiment news of unfavorable ratings
Enter in negative keyword set, is that the keyword occurred in the neutral public sentiment news evaluated is put into neutral keyword set by label.
Keyword in each keyword set is also possible to user and rule of thumb waits definition.
It is that the present invention is based on the processes of the public sentiment method for early warning 3rd embodiment of Recognition with Recurrent Neural Network algorithm with reference to Fig. 4, Fig. 4
Schematic diagram.
Based on above-mentioned second embodiment, the tendency degree of the keyword include positive tendency degree, negative tendency degree and in
Vertical tendency degree, the front tendency degree, negative tendency degree and neutral tendency degree respectively appear in positive new for the keyword
News, negative press, probability in neutral news, in the present embodiment, the step S101 is specifically included:
Step S1011: establishing keywords database, the keywords database include positive keyword set, negative keyword set and
Neutral keyword set；
It should be noted that the definition of keywords database can classify according to specific needs, it can be and closed including front
Keyword collection, negative keyword set and neutral keyword set are also possible to include positive keyword set, negative keyword set.
Step S1012: the correlation of each keyword with remaining keyword in each keyword set is calculated；
It should be noted that by the correlation for calculating each keyword and remaining keyword in each keyword set, it can
To determine the tendency degree of the keyword, such as keyword A, positive keyword set { A, B, C, D }, by the phase for calculating A and B, C, D
Guan Xinglai determines the positive tendency degree of A.
According to formulaRemaining calculated in the keyword and positive keyword set is crucial
The correlation of word；
Wherein, n is the number of keyword in positive keyword set；
Rec (w, v) is w, the correlation of two keywords of v；
P is positive keyword set；
P (w) is the probability that w keyword occurs in a document,
P (v) is the probability that v keyword occurs in a document；
P (w, v) is the probability that w and v occur in a document jointly.
Preferably, according to formulaIt calculates in the keyword and negative keyword set
The correlation of remaining keyword；
Wherein, m is the number of keyword in negative keyword set；
Rec (w, v) is w, the correlation of two keywords of v；
Q is positive keyword set；
P (w) is the probability that w keyword occurs in a document,
P (v) is the probability that v keyword occurs in a document；
P (w, v) is the probability that w and v occur in a document jointly.
Preferably, according to formulaIt calculates in the keyword and neutral keyword set
The correlation of remaining keyword；
Wherein, k is the number of keyword in neutral keyword set；
Rec (w, v) is w, the correlation of two keywords of v；
M is neutral keyword set；
P (w) is the probability that w keyword occurs in a document,
P (v) is the probability that v keyword occurs in a document；
P (w, v) is the probability that w and v occur in a document jointly.
Step S1013: according to the correlation of each keyword and remaining keyword in each keyword set, the key is calculated
Positive tendency degree, negative tendency degree and the neutral tendency degree of word.
When specific implementation, it can be the mean value of each keyword and the correlation of remaining keyword in each keyword set
Corresponding tendency degree as the keyword.
It is that the present invention is based on the processes of the public sentiment method for early warning fourth embodiment of Recognition with Recurrent Neural Network algorithm with reference to Fig. 5, Fig. 5
Schematic diagram.
Based on above-mentioned 3rd embodiment, in the present embodiment, the step S1013 is specifically included:
Step S1013a: correlation and negative keyword by keyword with remaining keyword in positive keyword set
The correlation for remaining keyword concentrated is inclined with the difference of correlation of remaining keyword in neutrality keyword set as front
Xiang Du；
It, can be with it should be noted that there is no precedence relationship between step S1013a, step S1013b and step S1013c
It is step S1013c preceding, step S1013a and step S1013b can also synchronize progress rear, therefore, do not do specific limit herein
System.
Remaining keyword when specific implementation, in positive tendency degree=keyword of a certain keyword and positive keyword set
Correlation-and remaining keyword in negative keyword set correlation-and remaining keyword in neutral keyword set
Correlation, i.e. rel1-rel2-rel3.
Usually, the phase of the positive tendency degree=keyword and remaining keyword in positive keyword set of a certain keyword
Close property average value-and the correlation of remaining keyword in negative keyword set average value-with neutrality keyword set in
The average value of the correlation of remaining keyword.
Step S1013b: by the correlation and front keyword of keyword and remaining keyword in negative keyword set
The correlation for remaining keyword concentrated negatively is inclined with the difference conduct of the correlation of remaining keyword in neutral keyword set
Xiang Du；
When specific implementation, the negative tendency degree=keyword and remaining keyword in negative keyword set of a certain keyword
Correlation-and remaining keyword in positive keyword set correlation-and remaining keyword in neutral keyword set
Correlation, i.e. rel2-rel1-rel3.
Usually, the phase of the negative tendency degree=keyword and remaining keyword in negative keyword set of a certain keyword
The average value-of the correlation of the average value-of closing property and remaining keyword in positive keyword set in neutrality keyword set
The average value of the correlation of remaining keyword.
Step S1013c: by the correlation and front keyword of keyword and remaining keyword in neutral keyword set
The correlation for remaining keyword concentrated is inclined with the difference of the correlation of remaining keyword in negative keyword set as neutrality
Xiang Du.
Remaining keyword when specific implementation, in neutral tendency degree=keyword of a certain keyword and neutral keyword set
Correlation-and remaining keyword in positive keyword set correlation-and remaining keyword in negative keyword set
Correlation, i.e. rel3-rel1-rel2.
Usually, the phase of the neutral tendency degree=keyword and remaining keyword in neutral keyword set of a certain keyword
The average value-of the correlation of the average value-of closing property and remaining keyword in positive keyword set in negative keyword set
The average value of the correlation of remaining keyword.
In addition, the embodiment of the present invention also proposes a kind of storage medium, it is stored on the storage medium based on circulation nerve
The public sentiment early warning program of network algorithm, it is real when the public sentiment early warning program based on Recognition with Recurrent Neural Network algorithm is executed by processor
The step of public sentiment method for early warning now as described above based on Recognition with Recurrent Neural Network algorithm.
It is that the present invention is based on the structures of the public sentiment prior-warning device first embodiment of Recognition with Recurrent Neural Network algorithm referring to Fig. 6, Fig. 6
Block diagram.
As shown in fig. 6, the public sentiment prior-warning device based on Recognition with Recurrent Neural Network algorithm that the embodiment of the present invention proposes includes:
Public sentiment obtains module 601, for obtaining public sentiment news in preset time, and determines keyword in the public sentiment news
Tendency degree；
It should be noted that public sentiment news is a kind of network public-opinion, it is diffused and is propagated by the network platform.Public sentiment is new
News can be through the publications such as webpage or third party software, plug-in unit.And the acquisition of public sentiment news can be and be connect by API
Mouth obtains, and is also possible to obtain by modes such as spiders, be not particularly limited herein.
The tendency degree of keyword can be divided into positive tendency degree, negative tendency degree, can also be divided into positive tendency degree, negatively
Tendency degree and neutral tendency degree.The positive tendency degree of keyword is the degree of front evaluation, and the negative tendency degree of keyword is
The degree of unfavorable ratings, the neutral tendency degree of keyword are the degree of neutral evaluation.
Vector establishes module 602, for the tendency degree according to the keyword, determine the corresponding feature of the keyword to
Amount；
It should be noted that the tendency degree according to the keyword, determines the corresponding feature vector of the keyword
It is to construct the corresponding feature vector of keyword using the tendency degree of keyword as corresponding weight.The dimension of feature vector can be with
Depending on the division of tendency degree, such as tendency degree is divided into positive tendency degree, negative tendency degree and neutral tendency degree, then feature
Vector can be set at least three-dimensional.
Sequence determining module 603, for determining the spy of the public sentiment news according to the corresponding feature vector of the keyword
Levy sequence；
It should be understood that public sentiment news be by multiple crucial phrases at, it is described according to the corresponding spy of the keyword
The step of levying vector, determining the characteristic sequence of the public sentiment news can be that the corresponding combination of eigenvectors of keyword is formed carriage
The characteristic sequence of feelings news.For example, keyword is m in a public sentiment news, the characteristic sequence for constructing the public sentiment news can be with
It, can also be according to the dimension of specific determining characteristic sequence of classifying for the dimensional feature vector of 3 × m or m × 3.
Index determining module 604, for the characteristic sequence of the public sentiment news to be inputted the Recognition with Recurrent Neural Network trained
Model determines public sentiment warning index；
It should be understood that using the characteristic sequence of public sentiment news as the input of Recognition with Recurrent Neural Network model, circulation nerve
The hidden layer of network includes the implicit vector of history public sentiment news, and by Recognition with Recurrent Neural Network model, the entirety for obtaining public sentiment is inclined
Xiang Du, as public sentiment warning index.
In addition, since public sentiment news is at a time or in the period, the regional range that public sentiment news is influenced is also
It is different, such as country, province, city etc., thus consider public sentiment news quantity be can be by the regional range of public sentiment news
As a correction value, it is denoted as t1.
Since the report of media also can have large effect to the quantity of public sentiment news, the exposure of public sentiment news
Degree is also considered as a correction value, is denoted as t2.
The circulation of public sentiment news also can reflect out public feelings information propagation condition and discussion temperature on network, because
This, the circulation of public sentiment news can also be used as a correction value, be denoted as t3.It is customized that the setting of correction value t3 can be user.
Such as the circulation possibility of the public sentiment news about the leaving office of company executives occurrences in human life might not be big, but for enterprise, this
The attention rate of one topic is often relatively high, therefore can be modified by adjusting t3.
Early warning issues module 605, for issuing public sentiment early warning according to the public sentiment warning index.
It is described according to the public sentiment warning index when specific implementation, public sentiment early warning is issued, can be in public sentiment warning index
When greater than a preset threshold, public sentiment early warning is issued.Wherein it is customized to can be user for preset threshold, can also be new according to public sentiment
The topic content of news is preset.The mode for issuing public sentiment early warning can be a variety of, such as pass through mobile portable phone, short message, mail
Etc. forms, be also possible to the customized simultaneous system of user.
The present invention determines the tendency degree of keyword in the public sentiment news by obtaining public sentiment news in preset time,
Further according to the tendency degree of the keyword, the corresponding feature vector of the keyword is determined, it is corresponding further according to the keyword
Feature vector determines the characteristic sequence of the public sentiment news, has finally trained the characteristic sequence input of the public sentiment news
Recognition with Recurrent Neural Network model determines public sentiment warning index, according to the public sentiment warning index, issues public sentiment early warning, can be accurate
Judge that public sentiment is moved towards, solves the technical problem of the prediction of the development trend effect difference of prior art public sentiment.
The present invention is based on the other embodiments of the public sentiment prior-warning device of Recognition with Recurrent Neural Network algorithm or specific implementation can
Referring to above-mentioned each method embodiment, details are not described herein again.
It should be noted that, in this document, the terms "include", "comprise" or its any other variant are intended to non-row
His property includes, so that the process, method, article or the system that include a series of elements not only include those elements, and
And further include other elements that are not explicitly listed, or further include for this process, method, article or system institute it is intrinsic
Element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that including being somebody's turn to do
There is also other identical elements in the process, method of element, article or system.
The serial number of the above embodiments of the invention is only for description, does not represent the advantages or disadvantages of the embodiments.
Through the above description of the embodiments, those skilled in the art can be understood that above-described embodiment side
Method can be realized by means of software and necessary general hardware platform, naturally it is also possible to by hardware, but in many cases
The former is more preferably embodiment.Based on this understanding, technical solution of the present invention substantially in other words does the prior art
The part contributed out can be embodied in the form of software products, which is stored in a storage medium
In (such as read-only memory/random access memory, magnetic disk, CD), including some instructions are used so that a terminal device (can
To be mobile phone, computer, server, air conditioner or the network equipment etc.) execute method described in each embodiment of the present invention.
The above is only a preferred embodiment of the present invention, is not intended to limit the scope of the invention, all to utilize this hair
Equivalent structure or equivalent flow shift made by bright specification and accompanying drawing content is applied directly or indirectly in other relevant skills
Art field, is included within the scope of the present invention.
Claims (10)
1. a kind of public sentiment method for early warning based on Recognition with Recurrent Neural Network algorithm, which comprises the steps of:
Public sentiment news in preset time is obtained, and determines the tendency degree of keyword in the public sentiment news；
According to the tendency degree of the keyword, the corresponding feature vector of the keyword is determined；
According to the corresponding feature vector of the keyword, the characteristic sequence of the public sentiment news is determined；
The characteristic sequence of the public sentiment news is inputted to the Recognition with Recurrent Neural Network model trained, determines public sentiment warning index；
According to the public sentiment warning index, public sentiment early warning is issued.
2. as described in claim 1 based on the public sentiment method for early warning of Recognition with Recurrent Neural Network algorithm, which is characterized in that the acquisition
Public sentiment news in preset time, and determine the step of tendency of keyword in the public sentiment news is spent, comprising:
Public sentiment news and the keywords database pre-established in preset time are obtained, and determines keyword in the public sentiment news
Tendency degree.
3. as claimed in claim 2 based on the public sentiment method for early warning of Recognition with Recurrent Neural Network algorithm, which is characterized in that the key
The tendency degree of word includes positive tendency degree, negative tendency degree and neutral tendency degree, the positive tendency degree, negative tendency degree with
And neutral tendency degree is that the keyword respectively appears in positive news, negative press, probability in neutrality news；
Correspondingly, public sentiment news and the keywords database pre-established in the acquisition preset time, and determine that the public sentiment is new
Before the step of tendency of keyword is spent in news, the public sentiment method for early warning based on Recognition with Recurrent Neural Network algorithm further includes as follows
Step:
Keywords database is established, the keywords database includes positive keyword set, negative keyword set and neutral keyword set；
Calculate the correlation of each keyword with remaining keyword in each keyword set；
According to the correlation of each keyword and remaining keyword in each keyword set, the front tendency of the keyword is calculated
Degree, negative tendency degree and neutral tendency degree.
4. as claimed in claim 3 based on the public sentiment method for early warning of Recognition with Recurrent Neural Network algorithm, which is characterized in that the basis
The correlation of each keyword and remaining keyword in each keyword set calculates the positive tendency degree of the keyword, negatively inclines
The step of Xiang Du and neutral tendency are spent, comprising:
By keyword and remaining keyword in the correlation and negative keyword set of remaining keyword in positive keyword set
Correlation, with the difference of the correlation of remaining keyword in neutral keyword set as positive tendency degree；
By keyword and remaining keyword in the correlation and positive keyword set of remaining keyword in negative keyword set
Correlation, with the difference of the correlation of remaining keyword in neutral keyword set as negative tendency degree；
By keyword and remaining keyword in the correlation and positive keyword set of remaining keyword in neutral keyword set
Correlation, with the difference of the correlation of remaining keyword in negative keyword set as neutral tendency degree.
5. as claimed in claim 3 based on the public sentiment method for early warning of Recognition with Recurrent Neural Network algorithm, which is characterized in that the calculating
The step of correlation of each keyword and remaining keyword in each keyword set, comprising:
According to formulaCalculate the keyword and remaining keyword in positive keyword set
Correlation；
Wherein, n is the number of keyword in positive keyword set；
Rec (w, v) is w, the correlation of two keywords of v；
P is positive keyword set；
P (w) is the probability that w keyword occurs in a document,
P (v) is the probability that v keyword occurs in a document；
P (w, v) is the probability that w and v occur in a document jointly.
6. as claimed in claim 3 based on the public sentiment method for early warning of Recognition with Recurrent Neural Network algorithm, which is characterized in that the calculating
The step of correlation of each keyword and remaining keyword in each keyword set, comprising:
According to formulaCalculate remaining keyword in the keyword and negative keyword set
Correlation；
Wherein, m is the number of keyword in negative keyword set；
Rec (w, v) is w, the correlation of two keywords of v；
Q is positive keyword set；
P (w) is the probability that w keyword occurs in a document,
P (v) is the probability that v keyword occurs in a document；
P (w, v) is the probability that w and v occur in a document jointly.
7. as claimed in claim 3 based on the public sentiment method for early warning of Recognition with Recurrent Neural Network algorithm, which is characterized in that the calculating
The step of correlation of each keyword and remaining keyword in each keyword set, comprising:
According to formulaCalculate the keyword and remaining keyword in neutral keyword set
Correlation；
Wherein, k is the number of keyword in neutral keyword set；
Rec (w, v) is w, the correlation of two keywords of v；
M is neutral keyword set；
P (w) is the probability that w keyword occurs in a document,
P (v) is the probability that v keyword occurs in a document；
P (w, v) is the probability that w and v occur in a document jointly.
8. a kind of public sentiment prior-warning device based on Recognition with Recurrent Neural Network algorithm characterized by comprising
Public sentiment obtains module, for obtaining public sentiment news in preset time, and determines the tendency of keyword in the public sentiment news
Degree；
Vector establishes module, for the tendency degree according to the keyword, determines the corresponding feature vector of the keyword；
Sequence determining module, for determining the characteristic sequence of the public sentiment news according to the corresponding feature vector of the keyword；
Index determining module, for the characteristic sequence of the public sentiment news to be inputted the Recognition with Recurrent Neural Network model trained, really
Determine public sentiment warning index；
Early warning issues module, for issuing public sentiment early warning according to the public sentiment warning index.
9. a kind of terminal, which is characterized in that the terminal includes: memory, processor and is stored on the memory and can
The public sentiment early warning program based on Recognition with Recurrent Neural Network algorithm run on the processor, it is described to be calculated based on Recognition with Recurrent Neural Network
The public sentiment early warning program of method be arranged for carrying out as described in any one of claims 1 to 7 based on Recognition with Recurrent Neural Network algorithm
The step of public sentiment method for early warning.
10. a kind of storage medium, which is characterized in that be stored with the public sentiment based on Recognition with Recurrent Neural Network algorithm on the storage medium
Early warning program realizes such as claim 1 when the public sentiment early warning program based on Recognition with Recurrent Neural Network algorithm is executed by processor
The step of to 7 described in any item public sentiment method for early warning based on Recognition with Recurrent Neural Network algorithm.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811530781.XA CN109800302A (en) | 2018-12-14 | 2018-12-14 | Public sentiment method for early warning, device, terminal and medium based on Recognition with Recurrent Neural Network algorithm |
| PCT/CN2019/122787 WO2020119533A1 (en) | 2018-12-14 | 2019-12-03 | Public sentiment warning method and apparatus based on recurrent neural network algorithm, terminal and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811530781.XA CN109800302A (en) | 2018-12-14 | 2018-12-14 | Public sentiment method for early warning, device, terminal and medium based on Recognition with Recurrent Neural Network algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109800302A true CN109800302A (en) | 2019-05-24 |
Family
ID=66556615
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811530781.XA Pending CN109800302A (en) | 2018-12-14 | 2018-12-14 | Public sentiment method for early warning, device, terminal and medium based on Recognition with Recurrent Neural Network algorithm |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN109800302A (en) |
| WO (1) | WO2020119533A1 (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020119533A1 (en) * | 2018-12-14 | 2020-06-18 | 深圳壹账通智能科技有限公司 | Public sentiment warning method and apparatus based on recurrent neural network algorithm, terminal and medium |
| CN112256974A (en) * | 2020-11-13 | 2021-01-22 | 泰康保险集团股份有限公司 | Public opinion information processing method and device |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20090048998A (en) * | 2007-11-12 | 2009-05-15 | 주식회사 비즈모델라인 | System and method for alarming bad public opinion using keyword and recording medium |
| US20130290232A1 (en) * | 2012-04-30 | 2013-10-31 | Mikalai Tsytsarau | Identifying news events that cause a shift in sentiment |
| CN104657393A (en) * | 2013-11-25 | 2015-05-27 | 深圳市至高通信技术发展有限公司 | Public opinion analysis method and corresponding device |
| CN108776671A (en) * | 2018-05-12 | 2018-11-09 | 苏州华必讯信息科技有限公司 | A kind of network public sentiment monitoring system and method |
| CN108959383A (en) * | 2018-05-31 | 2018-12-07 | 平安科技（深圳）有限公司 | Analysis method, device and the computer readable storage medium of network public-opinion |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR101074215B1 (en) * | 2009-02-27 | 2011-10-14 | 에스케이 텔레콤주식회사 | Apparatus for analyzing public opinion and Method for rating of public opinion through document analysis |
| CN105589941A (en) * | 2015-12-15 | 2016-05-18 | 北京百分点信息科技有限公司 | Emotional information detection method and apparatus for web text |
| CN107066442A (en) * | 2017-02-15 | 2017-08-18 | 阿里巴巴集团控股有限公司 | Detection method, device and the electronic equipment of mood value |
| CN109800302A (en) * | 2018-12-14 | 2019-05-24 | 深圳壹账通智能科技有限公司 | Public sentiment method for early warning, device, terminal and medium based on Recognition with Recurrent Neural Network algorithm |
-
2018
- 2018-12-14 CN CN201811530781.XA patent/CN109800302A/en active Pending
-
2019
- 2019-12-03 WO PCT/CN2019/122787 patent/WO2020119533A1/en active Application Filing
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20090048998A (en) * | 2007-11-12 | 2009-05-15 | 주식회사 비즈모델라인 | System and method for alarming bad public opinion using keyword and recording medium |
| US20130290232A1 (en) * | 2012-04-30 | 2013-10-31 | Mikalai Tsytsarau | Identifying news events that cause a shift in sentiment |
| CN104657393A (en) * | 2013-11-25 | 2015-05-27 | 深圳市至高通信技术发展有限公司 | Public opinion analysis method and corresponding device |
| CN108776671A (en) * | 2018-05-12 | 2018-11-09 | 苏州华必讯信息科技有限公司 | A kind of network public sentiment monitoring system and method |
| CN108959383A (en) * | 2018-05-31 | 2018-12-07 | 平安科技（深圳）有限公司 | Analysis method, device and the computer readable storage medium of network public-opinion |
Non-Patent Citations (1)
| Title |
|---|
| 尉译心: "情感倾向分析下的突发事件网络舆情预警分析", 《数字通信世界》, no. 07, pages 268 - 269 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020119533A1 (en) * | 2018-12-14 | 2020-06-18 | 深圳壹账通智能科技有限公司 | Public sentiment warning method and apparatus based on recurrent neural network algorithm, terminal and medium |
| CN112256974A (en) * | 2020-11-13 | 2021-01-22 | 泰康保险集团股份有限公司 | Public opinion information processing method and device |
| CN112256974B (en) * | 2020-11-13 | 2023-11-17 | 泰康保险集团股份有限公司 | Public opinion information processing method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2020119533A1 (en) | 2020-06-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2022161202A1 (en) | Multimedia resource classification model training method and multimedia resource recommendation method | |
| CN108804567A (en) | Improve method, equipment, storage medium and the device of intelligent customer service response rate | |
| US11556572B2 (en) | Systems and methods for coverage analysis of textual queries | |
| CN110110225B (en) | Online education recommendation model based on user behavior data analysis and construction method | |
| US8972404B1 (en) | Methods and systems for organizing content | |
| US20220172260A1 (en) | Method, apparatus, storage medium, and device for generating user profile | |
| CN109509010A (en) | A kind of method for processing multimedia information, terminal and storage medium | |
| CN106951471A (en) | A kind of construction method of the label prediction of the development trend model based on SVM | |
| CN107527240A (en) | A kind of operator's industry product Praise effect identification system and method | |
| CN110442872A (en) | A kind of text elements integrality checking method and device | |
| CN109214444B (en) | Game anti-addiction determination system and method based on twin neural network and GMM | |
| CN109800302A (en) | Public sentiment method for early warning, device, terminal and medium based on Recognition with Recurrent Neural Network algorithm | |
| Gao et al. | Application of artificial intelligence and big data technology in digital marketing | |
| CN112068962A (en) | Cloud rendering resource exchange method based on deep learning | |
| CN106649380A (en) | Hot spot recommendation method and system based on tag | |
| Lin | Learning information recommendation based on text vector model and support vector machine | |
| CN116957678A (en) | Data processing method and related device | |
| US20230085599A1 (en) | Method and device for training tag recommendation model, and method and device for obtaining tag | |
| CN115221954B (en) | User portrait method, device, electronic equipment and storage medium | |
| Chen | High-concurrency big data precision marketing and advertising recommendation under 5g wireless communication network environment | |
| Lian | Implementation of computer network user behavior forensic analysis system based on speech data system log | |
| US11610068B2 (en) | Systems and method for intent messaging | |
| CN111984842B (en) | Bank customer data processing method and device | |
| Xu et al. | Early warning of telecom customer churn based on multialgorithm model optimization | |
| Qu | Using data mining techniques to discover customer behavioural patterns for direct marketing |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |